The advantages of coding speech signal digitally are well known and are widely discussed in the literature [1]. Briefly, digital representation offers efficient signal regeneration, noise immunity, easy encryption, the possibility of combining transmission and switching functions, and the advantage of a uniform frmat for different types of signals. Unfortunately, these benefits are gained at the expense of increased transmission bandwidth. The redundancy removal systems (e.g., differential coding, linear prediction vocoders,…etc.) were developed to overcome this difficulty,, although, of the expense of system complexity and speech quality.
This paper introduces a simple adaptive differential pulse code modulation (ADPCM) system for speech coding at low bit rates. In this system line spectral pair (LSP) adaptive backward predictor is used to remove the redundancy present in the speech signal. Backward adaptation of the predictor coefficients is preferred due to the fact that it does not require a portion of the transmitted data rate to be allocated to the predictor coefficients, thus allowing the use of all bits available for coding the prediction residual (error). Furthermore, backward adaptation simplifies transmitter implementation.
Computer simulation experiments using Arabic speech bandllmited to 3.5KHz and sampled at 8 KHz, resulted in a high quality speech reproduction at bit rates between 24 - 32 Kbit/sec. Moreover, it is shown that the developed system performs well at bit error rate as high as 5%.